Effective segmentation for move objects in video

A novel scheme for automatic video segmentation is proposed in this paper by fully exploiting both inter-frame temporal information and intra-frame spatial information from the video sequence. The initial motion regions were detected in each frame by using a time-weighted queue. Many approximately homogenous color-texture regions were formed by using improved watershed based on a spatial similarity criterion and object size. Finally, further merging of temporal and spatial segment results by using a majority rule, and refining of them were performed based on spatial coherency and temporal continuity. As a result, the proposed method is robust and gets fine boundaries of objects. Furthermore, it can be adapted to high-speed and low-speed objects in indoor scene or outdoor scene by modifying a few parameters. The proposed spatial segment scheme can decrease the effect of the shadow as well.